A short presentation about, how to better design an AI Products using Product Thinking Principals meshed with AI Best Practice and learning from dealing with its Pitfalls.
Connect me at:
https://www.linkedin.com/in/saurabhkaushik
https://twitter.com/saurabhkaushik
2. About me (Professional Impact)
Improved user personalization to Find Realtor
through AI Intervention
Increased Audience engagement by measuring their
viewership on Digital and traditional channels.
Helped them migrate native applications on
Windows Embedded OS (CE)
Enabled AI in Touchless Account Payable
Enabled AI in Touchless Invoice to Cash
Enabled AI in Financial Controllership
Improved processing time and accuracy by
automating Financial Compliance process through AI
Built Media Optimization solution for Mobile
Devices
Built Contract and Product Entity Extraction
for Retail/Banking Clients
Built Customer Acquisition (Campaign
Management) Platform for Telco
Helped them innovate Smart Retail Banking
through AI
Built robust and scalable Private Cloud Platform
for their Fright Business
Faster processing and management of
Investment Managed Funds
Improved agility in their Drug Prescription and
Invoice Processing through Digital Interventions
Improved collaboration amongst various BUs of
Asset Management – Corporate Investment
Banking
Digital Transformation of Payment System -
Digital Banking
Build Remote Controlled Customer Care
Product for Broadband Telcom Customer
Automated Product Onboarding and Cataloging
Solution through AI Intervention
Improved collaboration and Compensation
processing with Insurance Brokers.
Faster processing of customer billing and invoicing
for Customers
Helped their customers to share media effortlessly
and view them elegantly on Mobile
Built Web Security Framework for Business
Portal
Built Personalization Product API and Helpdesk
Agent App for Zendesk.
3. Agenda
Not Artificial Intelligence (AI)
Not Tech Talk
Not Product Management
Not Product Thinking
Not Design Thinking
•Major Challenges with AI Product
•What is AI Product Thinking?
•Deeper dive in AI Product Thinking
pure AI Product Thinking
5. Why AI Products need different Product Thinking ?
Uber Self driving disaster
(Untrustworthy)
Microsoft Tay became racist
(Biased)
IBM Watson Oncology gave
bad recommendations
(Unexplainable)
6. Challenge: Dealing with real world data
Input
Output
Input
Output
Trained
Model
Explainable
Unbiased
Trustworthy
Unexplainable
Biased
Untrustworthy
Real World
Data
Traditional Products AI Products
Software
Programs
Logic
Processor
Algo
Processor
8. Challenge: Dealing with new Role between AI and Human
AI-first products are products that just would not
make sense without AI
AI-Inside Era AI-First EraPre-AI Era
Customer Service
10. Challenge: Designing UX for AI World
UX
Product
Technology
UX
Product + AI
Technology
UX
AI Product
Technology
Pre AI Era AI-Inside Era AI-First Era
User User User
11. Challenge: Bring back the focus on Product
UX / UI
Functional
Tech / AI Model
UX / UI
Functional
Tech / AI Model
AI Model is
the Product
This is the
Product
13. What is AI Product Thinking?
“Think in products, not in AI models”
Minus AI Engineering
Vision – Why are we doing this?
Strategy – How are we doing this?
Goals – What do we want to
achieve?
People – For whom are we
doing this?
Problem – What pain-point do
we solve?
Solution – What are we doing?
Users
14. How to do AI Product Thinking?
AI Product Thinking is a holistic approach of designing and developing Trustworthy, Unbiased and
Explainable AI Products by
Redefining new roles between AI and Human
Redesigning User Experience for AI and Edge Scenarios
Redrawing Testing mechanism to deal with real data
Reapplying Product Management with AI best Practices
17. User Role
• Product: Identify and deliver substantial value with which users will be comfortable
with letting a machine replace.
• UX: Identify and transition the change in most smooth, subtle and intuitive fashion.
Identify what users will STOP doing
• Product : Identify and renegotiate the deal between what humans to do and what
machines do.
• UX: Identify and maintain core features/controls to keen them in power seat.
Be very clear about what users will KEEP doing
Tech: Ensure, AI First to truly deliver new value with adequate levels of quality,
reliability and demonstrability.
26. Provide engineers with the right training data
Design to handle real-world situations in all layers of Product
Design to handle noise in data for consistent behavior
Design to handle biases in data to avoid embarrassing scenrios
Design to handle anomalies for edge cases
Design to handle acceptable accuracy in production over Product KPI
29. Vision - Visualizing the Future of Product
Product Vision for Customer (without
Tech/AI)
Reimagine your Product from AI
First/AI-inside era
Draw clear Value propositions
and differentiation
AI Product Vision and Mission
Statement
Product to help customer resolve their queries 24/7 with high degree of satisfaction
and quick resolution in seamless and effortless manner using AI-First Strategy.
Product to offer AI based Customer Assistant system to deliver Trustworthy, Sensitive,
Unbiased and Contextual query resolution through Conversational AI interface with
automated real time learning using power of Deep Learning tech to serve 10M+
queries per day.
High accuracy, relevancy, availability and automated learning loop at scale. Will
deliver on competitive Customer Facing KPI with a ChatBot who understand domain
best
To develop an AI based Customer Assistant system which will resolve customer
queries with high accuracy, relevancy, availability and automated learning loop at
scale to serve 10M queries per day for certain business domain – Retail Banking.
30. Strategy - Thinking about What to Build
Observe Product Trends with AI Impact
Follow latest and great innovation in AI
Develop AI Product Strategy and Roadmap
Build KPI matrices for overall Product and AI model
31. Design - Decide How to Build
Customer and Data Obsession in all decision making
Build Product with Simpler AI Algo first
Adapt Breath-First approach to build a product
Consider scalability and performance in Product Architecture
32. Execution - The Building Process
Follow Agile Development (Define/Validate/Iterate)
Ensure Product fails gracefully for edge conditions
Team Interaction: Understand fundamentals
Monitor Product Behavior and Customer Feedback
33. Interesting Quotes
To make human Interplanetary species - Elon Musk
“Fall in love with a problem, not a specific solution“ — Laura Javier
“Clean Thinking to Design Simple Product“ — Steve Jobs
"Focus on Product thinking, instead of feature or AI model" - Me ;-)